Advertisement

Gestational Choline Deficiency Causes Global and Igf2 Gene DNA Hypermethylation by Up-regulation of Dnmt1 Expression*

Open AccessPublished:August 27, 2007DOI:https://doi.org/10.1074/jbc.M705539200
      During gestation there is a high demand for the essential nutrient choline. Adult rats supplemented with choline during embryonic days (E) 11-17 have improved memory performance and do not exhibit age-related memory decline, whereas prenatally choline-deficient animals have memory deficits. Choline, via betaine, provides methyl groups for the production of S-adenosylmethionine, a substrate of DNA methyltransferases (DNMTs). We describe an apparently adaptive epigenomic response to varied gestational choline supply in rat fetal liver and brain. S-Adenosylmethionine levels increased in both organs of E17 fetuses whose mothers consumed a choline-supplemented diet. Surprisingly, global DNA methylation increased in choline-deficient animals, and this was accompanied by overexpression of Dnmt1 mRNA. Previous studies showed that the prenatal choline supply affects the expression of multiple genes, including insulin-like growth factor 2 (Igf2), whose expression is regulated in a DNA methylation-dependent manner. The differentially methylated region 2 of Igf2 was hypermethylated in the liver of E17 choline-deficient fetuses, and this as well as Igf2 mRNA levels correlated with the expression of Dnmt1 and with hypomethylation of a regulatory CpG within the Dnmt1 locus. Moreover, mRNA expression of brain and liver Dnmt3a and methyl CpG-binding domain 2 (Mbd2) protein as well as cerebral Dnmt3l was inversely correlated to the intake of choline. Thus, choline deficiency modulates fetal DNA methylation machinery in a complex fashion that includes hypomethylation of the regulatory CpGs within the Dnmt1 gene, leading to its overexpression and the resultant increased global and gene-specific (e.g. Igf2) DNA methylation. These epigenomic responses to gestational choline supply may initiate the long term developmental changes observed in rats exposed to varied choline intake in utero.
      An adequate supply of essential nutrients involved in the metabolism of methyl groups, including folic acid, vitamin B12, and choline, is central for normal development of the fetus. This is perhaps best exemplified by the discovery that the dietary supply of folic acid, a vitamin that acts as a coenzyme in one-carbon transfer pathways, during the periconceptual period is critical in prevention of neural tube defects (
      • Pitkin R.M.
      ). Studies in animal models (
      • Blusztajn J.K.
      ,
      • Meck W.H.
      • Kim C.L.
      ,
      • McCann J.C.
      • Kim M.
      • Ames B.N.
      ,
      • Zeisel S.H.
      ) as well as recent epidemiological investigations in humans (
      • Shaw G.M.
      • Kim S.L.
      • Yang W.
      • Selvin S.
      • Schaffer D.M.
      ) indicate that choline intake during gestation is particularly important for the normal development and function of the central nervous system. In a frequently used experimental model that employs offspring of pregnant rats or mice consuming diets of varying choline content during the 7-day period of the second half of gestation (embryonic days E11-17), prenatal choline deficiency causes deficits in certain memory tasks (
      • Meck W.H.
      • Kim C.L.
      ), whereas prenatal choline supplementation leads to enhanced memory and attention and prevents age-related memory decline (
      • Meck W.H.
      • Kim C.L.
      ,
      • Mellott T.J.
      • Kim C.L.
      • Meck W.H.
      • Blusztajn J.K.
      ,
      • Meck W.H.
      • Kim R.A.
      • Williams C.L.
      ,
      • Meck W.H.
      • Kim R.A.
      • Williams C.L.
      ,
      • Meck W.H.
      • Kim C.L.
      ,
      • Meck W.H.
      • Kim C.L.
      ,
      • Meck W.H.
      • Kim C.L.
      ). These behavioral changes are accompanied by electrophysiological, neuroanatomical, and neurochemical alterations that persist in the adulthood and old age (
      • Blusztajn J.K.
      ,
      • Meck W.H.
      • Kim C.L.
      ,
      • McCann J.C.
      • Kim M.
      • Ames B.N.
      ,
      • Zeisel S.H.
      ) and, remarkably, by altered patterns of brain gene expression postnatally (
      • Mellott T.J.
      • Kim M.T.
      • Diesl V.
      • Hill A.A.
      • Lopez-Coviella I.
      • Blusztajn J.K.
      ). Although the molecular mechanisms for these long-term effects of prenatal choline intake are not known, it is possible that they are related to the function of choline as a methyl group donor in one of the two alternative pathways that produce methionine by scavenging homocysteine. The latter compound is derived from S-adenosylhomocysteine (SAH),
      The abbreviations used are: SAH, S-adenosylhomocysteine; SAM, S-adenosylmethionine; DNMTs, DNA methyltransferases; DMD, differentially methylated domain; HPLC, high performance liquid chromatography; RT, reverse transcription; DMR2, differentially methylated region 2; ANOVA, analysis of variance; Sup, choline supplemented; Con, control; Def, choline-deficient.
      2The abbreviations used are: SAH, S-adenosylhomocysteine; SAM, S-adenosylmethionine; DNMTs, DNA methyltransferases; DMD, differentially methylated domain; HPLC, high performance liquid chromatography; RT, reverse transcription; DMR2, differentially methylated region 2; ANOVA, analysis of variance; Sup, choline supplemented; Con, control; Def, choline-deficient.
      a metabolite of the “universal” methyl donor, S-adenosylmethionine (SAM). The main pathway that regenerates methionine from homocysteine uses methyltetrahydrofolate. An alternative pathway uses betaine, the product of enzymatic oxidation of choline (
      • Finkelstein J.D.
      • Kim J.J.
      • Harris B.J.
      • Kyle W.E.
      ). Further downstream, methionine is used to generate SAM. After SAM donates its methyl group through various methylation reactions in the cell, its metabolite SAH enters the cycle again. Thus, SAM and SAH are indicators of the methylation capacity of the cell. Because in mammals one-carbon metabolism is dependent on the dietary supply of methyl group donors and cofactors (
      • Van den Veyver I.B.
      ), folate, methionine, vitamin B12, and choline can influence levels of SAM and SAH in the cell. Indeed, rats ingesting a choline-deficient diet have diminished tissue concentrations of methionine and SAM (
      • Poirier L.A.
      • Kim P.H.
      • Rogers A.E.
      ,
      • Shivapurkar N.
      • Kim L.A.
      ,
      • Zeisel S.H.
      • Kim T.
      • DaCosta K.-A.
      • Pomfret E.A.
      ). Because SAM is a donor of methyl groups for most biological methylations, SAM availability can affect DNA methylation, a covalent modification of cytosines in the CpG dinucleotide context by the addition of a methyl group to the 5-position. DNA methylation is a major epigenetic mechanism that mediates genomic imprinting and X-chromosome inactivation and regulates tissue-specific gene expression, cell differentiation, and chromatin structure (
      • Bird A.
      ,
      • Bernstein B.E.
      • Kim A.
      • Lander E.S.
      ). The reported effects of choline or methyl group deficiency on global DNA methylation vary; most authors observed hypomethylation (
      • Alonso-Aperte E.
      • Kim G.
      ,
      • Wainfan E.
      • Kim M.
      • Stender M.
      • Christman J.K.
      ,
      • Wilson M.J.
      • Kim N.
      • Poirier L.A.
      ,
      • Christman J.K.
      • Kim G.
      • Dizik M.
      • Abileah S.
      • Wainfan E.
      ,
      • Niculescu M.D.
      • Kim C.N.
      • Zeisel S.H.
      ,
      • Bhave M.R.
      • Kim M.J.
      • Poirier L.A.
      ,
      • James S.J.
      • Kim I.P.
      • Pogribna M.
      • Miller B.J.
      • Jernigan S.
      • Melnyk S.
      ,
      • Kim Y.I.
      • Kim I.P.
      • Basnakian A.G.
      • Miller J.W.
      • Selhub J.
      • James S.J.
      • Mason J.B.
      ), some no change (
      • Kim Y.I.
      • Kim J.K.
      • Fleet J.C.
      • Cravo M.L.
      • Salomon R.N.
      • Smith D.
      • Ordovas J.
      • Selhub J.
      • Mason J.B.
      ), and some hypermethylation (
      • Song J.
      • Kim K.J.
      • Medline A.
      • Ash C.
      • Gallinger S.
      • Kim Y.I.
      ,
      • Sohn K.J.
      • Kim J.M.
      • Reid S.
      • Shirwadkar S.
      • Mason J.B.
      • Kim Y.I.
      ), suggesting that DNA methylation may respond to the supply of methyl groups in a complex fashion that includes alterations in the activities of DNA methylating and/or demethylating enzymes. DNA methylation patterns acquired during development may be inherited through the cell divisions in a process catalyzed by DNA methyltransferase 1 (DNMT1) that methylates hemimethylated CpG sites and, thus, restores the parental methylation pattern on the daughter DNA strand after DNA replication (
      • Hsu D.W.
      • Kim M.J.
      • Lee T.L.
      • Wen S.C.
      • Chen X.
      • Shen C.K.
      ). In addition, DNMT3A and DNMT3B generate the DNA methylation patterns de novo during development and in adulthood (
      • Okano M.
      • Kim E.
      ,
      • Singal R.
      • Kim J.M.
      ). Diet can also influence the expression of DNMTs. Fischer male rats fed a methyl group deficient diet for at least 3 weeks display global DNA hypomethylation and, perhaps paradoxically, increased expression and activity of cellular DNMTs, possibly as compensatory mechanism (
      • James S.J.
      • Kim I.P.
      • Pogribna M.
      • Miller B.J.
      • Jernigan S.
      • Melnyk S.
      ,
      • Ghoshal K.
      • Kim X.
      • Datta J.
      • Bai S.
      • Pogribny I.
      • Pogribny M.
      • Huang Y.
      • Young D.
      • Jacob S.T.
      ). DNA methylation within regulatory regions of many genes alters their expression. In most cases, this results in gene silencing (
      • Nan X.
      • Kim H.H.
      • Johnson C.A.
      • Laherty C.D.
      • Turner B.M.
      • Eisenman R.N.
      • Bird A.
      ,
      • Jones P.L.
      • Kim G.J.
      • Wade P.A.
      • Vermaak D.
      • Kass S.U.
      • Landsberger N.
      • Strouboulis J.
      • Wolffe A.P.
      ,
      • Eden S.
      • Kim T.
      • Keshet I.
      • Cedar H.
      • Thorne A.W.
      ,
      • Takizawa T.
      • Kim K.
      • Namihira M.
      • Ochiai W.
      • Uemura A.
      • Yanagisawa M.
      • Fujita N.
      • Nakao M.
      • Taga T.
      ). In some instances, DNA methylation can enhance gene expression, such as in the insulin-like growth factor (Igf2) gene, in which CpG methylation of regulatory regions prevents binding of repressors to silencer elements (
      • Murrell A.
      • Kim S.
      • Bowden L.
      • Constancia M.
      • Dean W.
      • Kelsey G.
      • Reik W.
      ,
      • Eden S.
      • Kim M.
      • Hashimshony T.
      • Dean W.
      • Goldstein B.
      • Johnson A.C.
      • Keshet I.
      • Reik W.
      • Cedar H.
      ). There are several genomic regions within Igf2 that control its expression in a methylation-dependent manner, including the differentially methylated domain (DMD) located upstream of the H19 promoter that functions as a methylation-dependent insulator (
      • Bell A.C.
      • Kim G.
      ) and the differentially methylated region 2 (DMR2), located within the last exon of the peptide coding region of the gene (
      • Murrell A.
      • Kim S.
      • Bowden L.
      • Constancia M.
      • Dean W.
      • Kelsey G.
      • Reik W.
      ). When DMR2 is methylated, Igf2 expression is increased (
      • Murrell A.
      • Kim S.
      • Bowden L.
      • Constancia M.
      • Dean W.
      • Kelsey G.
      • Reik W.
      ). Igf2 is highly expressed in fetal liver, muscle, and choroid plexus (
      • Bondy C.A.
      • Kim H.
      • Roberts Jr., C.T.
      • LeRoith D.
      ). We found that Igf2 is overexpressed in brain of prenatally choline-supplemented adult rats and that its expression is reduced by prenatal choline deficiency (
      • Mellott T.J.
      • Kim M.T.
      • Diesl V.
      • Hill A.A.
      • Lopez-Coviella I.
      • Blusztajn J.K.
      ).
      We examined the components of the DNA methylating machinery in E17 rat embryos derived from mothers consuming varying amounts of choline. SAM levels were increased in liver and brain of the choline-supplemented animals as compared with controls. Surprisingly, choline-deficient embryos had global and Igf2 DMR2 DNA hypermethylation in liver that was accompanied by reduced levels of methylation in a regulatory CpG within the Dnmt1 gene concomitant with the induction of Dnmt1 expression. The data suggest that maternal choline deficiency causes an apparently compensatory induction of Dnmt1 expression in the fetus that prevents the loss of DNA methylation when choline (and possibly other sources of metabolic methyl groups) are in short supply.

      EXPERIMENTAL PROCEDURES

      Dietary Intervention during Embryonic Period E11-E17—Several cohorts of pregnant Sprague-Dawley rats (CD strain, Charles River Laboratories) were divided into three groups of four animals per experiment and fed either a choline-supplemented, control, or choline-deficient diet during E11-E17 of gestation (AIN-76A Rodent Purified Diet (
      • Bieri J.G.
      • Kim G.S.
      • Briggs G.M.
      • Phillips R.W.
      • Woodard J.C.
      • Kanapka J.J.
      ,
      • Bieri J.G.
      ) (Dyets Inc., Bethlehem, PA). Unlike commercial rat chows whose choline content is not controlled for, the AIN-76A diet was formulated to permit standardized studies using nutritionally adequate diet in rats and mice (
      • Bieri J.G.
      • Kim G.S.
      • Briggs G.M.
      • Phillips R.W.
      • Woodard J.C.
      • Kanapka J.J.
      ,
      • Bieri J.G.
      ). Choline was supplied in the food (as choline chloride) such that the choline-supplemented diet had 36 mmol/kg choline, the control diet had 8 mmol/kg choline, and the deficient diet had 0 mmol/kg choline. The consumption of the choline-deficient diet during this period of pregnancy causes a >50% reduction of the maternal choline and phosphocholine pools (
      • Zeisel S.H.
      • Kim M.H.
      • Zhou Z.W.
      • Da Costa K.A.
      ). The animals were housed individually with a 12-h light/dark cycle. The average litter size was not influenced by the diet (deficient 12.4, control 12.1, supplemented 12.1 fetuses per dam). All animal procedures were performed in accordance with protocols approved by the Boston University School of Medicine Institutional Animal Care and Use Committees.
      Dissection of Embryonic Liver and Frontal Cortex—The dams were euthanized on E17 after anesthesia with CO2, and the uteri were rapidly dissected and placed in L15 medium (Invitrogen) on ice. Subsequently the liver and the whole brain (for the determination of SAM and SAH and for global DNA methylation analysis) were dissected keeping the tissue on ice to minimize postmortem changes in the metabolite levels. There was no effect of the diet on the average weight of the liver (deficient 38 mg; control 37 mg; supplemented 37 mg) or brain (deficient 51 mg; control 52 mg; supplemented 51 mg). In addition, the fronto-parietal cortex (for all other assays) was similarly dissected. At this embryonic age, the frontal cortex is not fully differentiated, and to avoid inconsistency, the whole fronto-parietal cortex was taken making sure to exclude the olfactory bulbs and remove the meninges. A dissecting stereo microscope was used to aid in the preparation of the brain. The cortices from four embryos obtained from one mother were pooled, and those pools were used in all subsequent experiments. For liver dissection, a part of the right lobe was collected. The livers from eight embryos obtained from one mother were pooled, and those pools were used in all subsequent experiments.
      Genomic DNA, Total RNA, and Protein Isolation—For DNA, the tissue was immediately frozen on dry ice and stored at -70 °C until extraction. To obtain high yield pure genomic DNA from liver and whole brain, the Genomic tip 100/G kit (Qiagen) was used, and the protocol of the manufacturer was followed precisely. For frontal cortex, a DNeasy Blood and Tissue kit (Qiagen) was used. The concentration of DNA was determined via spectrophotometry at 260 nm.
      For RNA, the tissue was promptly homogenized using a needle and syringe in ice-cold 4 m guanidine isothiocyanate solution, pH 7.0 (containing 100 mm β-mercaptoethanol and 25 mm sodium citrate), and placed on dry ice. The samples were stored at -70 °C. The extraction was performed according to the phenol/chloroform method (
      • Chomczynski P.
      • Kim N.
      ), and the RNA was precipitated with ethanol and resuspended in nuclease-free water (Ambion). The final concentration was determined by fluorometry using the RiboGreen® RNA Quantitation reagent and kit (Molecular Probes) which allows precise quantitation in the nanogram range. The results were obtained using Victor3 multilabel plate reader (PerkinElmer Life Sciences).
      Determination of SAM and SAH—Tissue concentrations of SAM and SAH were measured by HPLC using a modification of previously described methods (
      • Shivapurkar N.
      • Kim L.A.
      ,
      • Molloy A.M.
      • Kim D.G.
      • Kennedy G.
      • Kennedy S.
      • Scott J.M.
      ). Briefly, 50 mg of tissue was homogenized in 185 μl of 40% trichloroacetic acid and 250 μl of cold 0.1 m sodium acetate, pH 6.0. After centrifugation (5 min, 9000 × g at 4 °C), the supernatant was washed 3 times with petroleum ether. The samples were centrifuged (5 min, 14,000 × g); the supernatant was transferred to another tube; and 100 μl of the sample was applied to Beckman Ultrasphere column (25 cm long, 5-μm porosity). The elution was performed at a rate of 1.6 ml/min with a two buffer system consisting of buffer A (25 mm sodium phosphate, 10 mm 1-heptanesulfonic acid, pH 3.2) and buffer B (80% v/v methanol, 25 mm sodium phosphate, 10 mm 1-heptanesulfonic acid, pH 3.2). The elution was isocratic at 0% buffer B for 4 min, and then a stepwise gradient of mobile phase was used to 5% B over 1 min that was maintained for 20 min. This was followed by a rapid change to 100% B over 1 min that was maintained for 10 min and, finally, a rapid re-equilibration to 0% B over 1 min and a subsequent 15-min period of 0% B. SAM and SAH were detected by UV absorbance (254 nm) and identified by comparing retention times with those of authentic standards (Sigma), and their amounts were quantified using a standard curve in the range of 125 pmol to 1500 pmol of each SAM and SAH.
      Analysis of 5-Methylcytosine DNA Content—The 5-methylcytosine content was determined using a modified procedure described by Gehrke et al. (
      • Gehrke C.W.
      • Kim R.A.
      • Gama-Sosa M.A.
      • Ehrlich M.
      • Kuo K.C.
      ). Briefly, 20 μg of total DNA from liver and brain was denatured by heating to 100 °C for 5 min and then hydrolyzed with nuclease P1 (Sigma) and subsequently with alkaline phosphatase (Sigma). The nucleosides were separated with HPLC with Supelcosil LC18 column (25 cm long, 5-μm porosity). We used a two buffer system that consisted of buffer A (2.5% v/v methanol, 50 mm potassium phosphate, pH 4.5) and buffer B (16% v/v methanol, 50 mm potassium phosphate, pH 4.5). The elution was isocratic at 0% buffer B for 16 min, and then a multi-segment linear gradient of mobile phase was used to 3.5% B over 2 min, to 8.5% B over 4 min, to 16% B over 6 min, to 22% B over 2 min, to 31% B over 6 min, and to 100% B over 4 min. One hundred percent of buffer B elution was maintained for an additional 10 min followed by a rapid reequilibration to 0% B over 1 min and a subsequent 14 min period of 0% B. The flow rate was 1 ml/min, and the cytidine nucleosides were detected by absorbance at 280 nm. At this wavelength the detection is more sensitive than at the usual 254 nm because cytosine and 5-methylcytosine have a high coefficient of absorbance at 280 nm. To identify the nucleosides, the retention times were compared with those of standard 2′-deoxycytidine (Sigma) and 5-methyl-2′-deoxycytidine (Berry & Associates). Global methylation was calculated as the ratio between the height of the 5-methyldeoxycytidine peak over the sum of the heights of 5-methyl- and deoxycytidine peaks.
      Semiquantitative Reverse Transcriptase-PCR Analysis—Total RNA from liver and frontal cortex was extracted using the phenol/chloroform method (
      • Chomczynski P.
      • Kim N.
      ), and the RNA was precipitated with ethanol and resuspended in nuclease-free water (Ambion). The final RNA concentration was determined by fluorometry using the RiboGreen® RNA Quantitation reagent and kit (Molecular Probes). Total RNA (15-30 ng) was amplified using the Superscript™ One-Step RT-PCR with Platinum Taq (Invitrogen) according to the manufacturer's instructions. The following primers were used: Dnmt1, Dnmt3a, Dnmt3b (
      • Pruitt K.
      • Kim A.S.
      • Frantz K.
      • Rojas R.J.
      • Muniz-Medina V.M.
      • Rangnekar V.M.
      • Der C.J.
      • Shields J.M.
      ), Mbd2 (
      • Esfandiari F.
      • Kim R.
      • Cotterman R.F.
      • Pogribny I.P.
      • James S.J.
      • Miller J.W.
      ), β-actin, and Igf2 (
      • Mellott T.J.
      • Kim M.T.
      • Diesl V.
      • Hill A.A.
      • Lopez-Coviella I.
      • Blusztajn J.K.
      ). The primers had the following sequence: Dnmt1, forward CCA GAT ACC TAC CGG TTA TTC G, reverse TCC TTT AAC TGC AGC TGA GGC; Dnmt3a, forward CTG AAA TGG AAA GGG TGT TTG GC, reverse CCA TGT CCC TTA CAC ACA AGC; Dnmt3b, forward GTA CTT CTG GGG TAA CCT ACC, reverse GCA AAC AGG TGT CTG ATG ACC; Dnmt3l forward AAG ACC CAT GAA ACC TTG AAC C, reverse, GTT GAC TTC GTA CCT GAT GAC CTC; Mbd2, forward GGC AAG AGC GAT GTC TAC TA, reverse CTG GAC CGA CTC CTT GAA GA; β-actin, forward, CAC AGC TGA GAG GGA AAT C, reverse TCA GCA ATG CCT GGG TAC; Igf2, forward, CCAGGT GAC AGG ACT GGC AC, reverse, CCT GAA AAC ACC CAT CCC AC. The cycling conditions were as follows: 1 cycle at 48 °C for 45 min; various numbers of cycles of denaturing at 94 °C for 1 min; annealing at 58 °C (for Dnmt3a 60 °C was used, and for Dnmt3b 62 °C was used) for 1 min and extension 70 °C for 2 min; 1 cycle of 72 °C for 7 min and hold at 4 °C. We used 15 ng of template for all reactions except for Dnmt3l and Igf2, for which 30 ng were used. The number of cycles used were as follows: β-actin, 30 cycles, Dnmt1, 32 cycles; Dnmt3a and Dnmt3b, 34 cycles; Mbd2, Dnmt3l, 35 cycles in liver and 38 cycles in cortex. The conditions of the reactions were determined to be in the linear range for each RNA species. The products were resolved on 10% Tris-buffered EDTA polyacrylamide gels and stained with ethidium bromide. The intensity of each band was quantified with a Kodak Image Station using Kodak ID software. The expression level was calculated as percent of the control after normalizing to β-actin.
      Bisulfite Sequencing of Igf2 Gene—Bisulfite sequencing is used to determine the methylation pattern of relatively large (up to 700 bp) regions of DNA with a single CpG resolution (
      • Grunau C.
      • Kim S.J.
      • Rosenthal A.
      ). All protocols were based on the method developed by Frommer et al. (
      • Frommer M.
      • Kim L.E.
      • Millar D.S.
      • Collis C.M.
      • Watt F.
      • Grigg G.W.
      • Molloy P.L.
      • Paul C.L.
      ). Genomic DNA (1 μg) from liver and frontal cortex was bisulfite-treated using the EZ DNA methylation kit™ (Zymo Research). The protocol of the manufacturer was followed precisely except for the conversion step, which was done at 55 °C for 16 h because this modification produced almost 100% conversion of non-methylated cytosines. The total yield was determined using the RiboGreen® RNA quantitation reagent and kit (Molecular Probes). Although RiboGreen reagent is designed for RNA quantification, we determined that this reagent can precisely and reliably quantify bisulfite-treated DNA because it becomes largely single stranded as the vast majority of cytosines are deaminated. Bisulfite-treated DNA was stored at -20 °C, and all excess pipetting and vortexing was avoided to prevent shearing of the DNA. Next, 20-100 ng of the template was amplified by PCR with primers specifically designed to amplify only converted template regardless of the methylation status. All primers for bisulfite PCR were designed using the MethPrimer software (
      • Li L.C.
      • Kim R.
      ). The primers were designed so that they amplify the core of DMR2 and DMD r3 and had the following sequence: DMR2 forward, TAT TGA TAT TTG GAT GGG AGT TTA G, and reverse, TAA AAC TAT CCC TAC TCA AAA AAA A; DMD r3 forward, TTG AGT GGA AAG ATT AAA ATT GTT A, and reverse, CCC CAA AAT TAA CTA CAT CTA AAC TAC. The 25-μl reaction contained template, 2 μl of primer mix (5 μm each), 2.5 μl of 10× high fidelity PCR buffer, 1 μl of mm MgSO4, 0.2 μl of Platinum® Taq High Fidelity (Invitrogen), 1 μl of 10 mm dNTP mixture, and nuclease-free water (Ambion). The reaction conditions were: 1 cycle at 94 °C for 2 min, 55 °C for 1 min, and 70 °C for 2 min; 30 cycles (the same number of cycles was used for all primers) of denaturing at 94 °C for 30 s, annealing at 55 °C for 40 s, and extension at 70 °C for 1 min; 1 cycle of 72 °C for 10 min and hold at 4 °C. Promptly after PCR, 10 μl of the reaction was resolved on 1% agarose gel to verify the size of the product and determine the approximate concentration. The remaining amount of the PCR reaction was used for subsequent cloning with the TA cloning® kit (Invitrogen). Generally, 5-7 ng of PCR product were ligated into the pCR®2.1 vector and used for transformation of competent Escherichia coli. The cells were plated on LB agar containing 100 μg/ml ampicillin to select for the cells that obtained the plasmid. Because the PCR product was inserted in the vector within the lacZα gene, the plates had 40 μl of 40 mg/ml X-gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) to help distinguish the colonies that obtained the plasmid with the ligated PCR product (the color is white) from the colonies that obtained only empty plasmid (blue color). To analyze the transformants, 10-30 colonies were selected and grown overnight in 5 ml of LB broth containing 100 μg/ml ampicillin. The plasmid was purified using the QIAprep® Spin Miniprep kit (Qiagen), and 10-25 clones per animal, 4 animals per treatment group were sequenced at the Massachusetts General Hospital DNA Sequencing Core. For the whole cloning experiment, a total of 2500 clones were sequenced. The results from DMR2 were analyzed using BiQ Analyzer (
      • Bock C.
      • Kim S.
      • Mikeska T.
      • Paulsen M.
      • Walter J.
      • Lengauer T.
      ), a computer program that selects for analysis only the clones that display a 95% conversion of C to T (indicating high efficiency of bisulfite-mediated deamination of cytosines) and that are different from each other, i.e. originate from separate DNA templates. For DMD, the results were analyzed manually.
      Methylation-specific PCR for Dnmt1 Gene—Methylation-specific PCR is a PCR-based method that permits the discrimination of methylated from unmethylated template. The assay utilizes bisulfite-converted DNA and primers specific for methylated and unmethylated DNA. Methylation-specific PCR was performed using DNA treated with bisulfite as described above. The primers were designed using the MethPrimer software (
      • Li L.C.
      • Kim R.
      ). Because the purpose of the experiment was to assay a single CpG that is thought to regulate Dnmt1 expression, the forward primer was designed for methylated or unmethylated template containing the CpG of interest at the 3′ end, and the reverse primer was common and wobbled in order not to be methylation-specific. All primers were tested with different annealing temperatures, and the highest effective temperature was chosen to avoid PCR bias (
      • Shen L.
      • Kim Y.
      • Chen X.
      • Ahmed S.
      • Issa J.P.
      ). The primer sequences are: Dnmt1 methylated forward, GTT GGA ATT TTA TTT TTG AGT TGC; unmethylated forward, TTG GAA TTT TAT TTT TGA GTT GT; common reverse, TAT AAC TAA CCR CCT TTC ACT AAC R (R = A or G). The reaction mixture and the reaction conditions were similar to the ones described above. All reactions were performed with 30 ng of template; 30-35 cycles. The annealing temperatures were: Dnmt1 methylated, 56.9 °C; Dnmt1 unmethylated, 55.2 °C. The conditions of the reactions were determined to be in the linear range of the assay. The products were resolved on 10% Tris-buffered EDTA polyacrylamide gels and stained with ethidium bromide. The intensity of each band was quantified with a Kodak Image Station using Kodak ID software. The methylation level was calculated as the ratio of the intensity of the methylated product divided by the unmethylated product.
      Statistical Analysis—The data were analyzed by a one or two way ANOVA as appropriate. The significant effects (p < 0.05) were further analyzed by multiple comparison tests such as the Tukey test. Microsoft Excel and SYSTAT (Systat Software Inc.) software for Macintosh computers were used.

      RESULTS

      SAM Levels in Liver and Brain in E17 Embryos—To determine the effect of maternal choline intake on fetal methylation capacity, we first measured hepatic and brain SAM and SAH levels. E17 embryos derived from pregnant rats fed a choline-supplemented diet during E11-17 had an average SAM content in liver of 55.1 ± 5.0 nmol/g (Fig. 1) and 43.0 ± 2.7 nmol/g in brain (Fig. 1). Both of these values were higher as compared with the SAM content of embryos from rats on control diet (p < 0.05) by 17%. In contrast, choline deficiency had no detectable effect on SAM levels in liver and brain as compared with control and choline-supplemented animals. Moreover, maternal choline intake did not affect the levels of SAH in liver (deficient 1.97 ± 0.2; control 1.67 ± 0.07; supplemented 1.74 ± 0.2 nmol/g) and brain (deficient 0.61±.06; control 0.48 ± 0.06; supplemented 0.50 ± 0.06 nmol/g), and there was no change between the three groups in the SAM/SAH ratio, a marker of the cellular methylation capacity (data not shown).
      Figure thumbnail gr1
      FIGURE 1SAM levels in the liver and brain of E17 embryos. Four pregnant dams per dietary group were used, and the tissues of four embryos from each dam were pooled for analysis. Each pool constituted an experimental sample. SAM levels were determined using HPLC as described under “Experimental Procedures.” Data are presented as the means ± S.E. In both liver and brain, choline-supplemented embryos had higher SAM levels as compared with control embryos (p < 0.05; Tukey test).
      Global DNA Methylation in E17 Liver and Brain—We next examined the 5-methylcytosine contents in DNA using HPLC. Surprisingly, in the liver of choline-deficient embryos, 5-methylcytosine content of DNA was significantly higher than that of controls (4.1 ± 0.1 mol% versus 3.8 ± 0.1 mol%) (Fig. 2), and the DNA from the choline-supplemented brains was hypomethylated (Fig. 2), with a 5-methylcytosine content of 4.3 ± 0.1 mol%, as compared with the control and choline-deficient with 5-methylcytosine contents of 5.1 ± 0.2 and 4.8 ± 0.2 mol %, respectively. Even though the differences between the groups for both liver and brain were small, they were statistically significant (p < 0.05). Note that only a small proportion of the total DNA cytosine occurs in the CpG context that constitutes the substrate for DNA methyltransferases, and thus, even large changes in the CpG methylation status would represent only a small fraction of the total cytosine. Interestingly, the average level of methylation was higher in the brain as compared with liver. These results indicate that maternal choline intake affects global DNA methylation in fetal liver and brain. Surprisingly, after a 7-day midgestation alteration in choline availability, there was an apparently inverse relationship between choline intake and DNA methylation.
      Figure thumbnail gr2
      FIGURE 25-Methylcytosine content in genomic DNA in the liver and brain of E17 rat embryos. DNA was purified and digested by nuclease P1 and alkaline phosphatase, and the resulting nucleosides were analyzed by HPLC as described under “Experimental Procedures.” Data are presented as the means ± S.E. of four embryo pools as described in . In the liver the choline-deficient embryos had higher global DNA methylation as compared with control embryos (p < 0.05; Tukey test). Similar results were obtained in a separate experiment using an additional four subjects per group (data not shown). In the brain choline-supplemented embryos had hypomethylated DNA as compared with controls (p < 0.05; Tukey test).
      Igf2 DMR2 Methylation Status in E17 Liver—Analysis of global DNA methylation described above provides information on the degree of DNA methylation that occurs primarily in the repetitive, non-coding, and non-regulatory sequences, and thus, we wished to determine whether choline availability also affects methylation of DNA regions known to regulate gene expression in a methylation-dependent manner. We elected to study regulatory sequences of Igf2, a gene whose regulation by DNA methylation is well studied. We focused on the differentially methylated region 2 (DMR2) because in contrast to other DMRs of the Igf2 locus, the DMR2 methylation signature changes dramatically during development (
      • Lopes S.
      • Kim A.
      • Hajkova P.
      • Dean W.
      • Oswald J.
      • Forne T.
      • Murrell A.
      • Constancia M.
      • Bartolomei M.
      • Walter J.
      • Reik W.
      ). In gametes DMR2 has a specific methylation pattern that becomes erased by active demethylation after fertilization. A new pattern becomes re-established over a relatively long time period and is nearly complete in the late fetal stages. Thus, this prolonged period may render DMR2 susceptible to environmental, including nutritional, changes. Using bisulfite PCR, cloning, and sequencing of DNA isolated from E17 liver and frontal cortex, we determined the methylation status of the 380-bp region in the core of DMR2 that contained 20 CpGs. This region of the rat genome (Ensembl genomic location chr 1:202909439-202909818) corresponds to the well studied murine gene (Ensembl genomic location chr 7:142463134-142463509) (
      • Murrell A.
      • Kim S.
      • Bowden L.
      • Constancia M.
      • Dean W.
      • Kelsey G.
      • Reik W.
      ). We used arbitrary numbering of the CpGs such that CpG numbered 5-12 correspond to the region investigated by Murrell et al. (
      • Murrell A.
      • Kim S.
      • Bowden L.
      • Constancia M.
      • Dean W.
      • Kelsey G.
      • Reik W.
      ). Most clones were partially methylated, and only a small fraction of the clones was fully unmethylated. The percentage of methylation per dietary group was calculated as the average number of clones for each of the 20 CpGs that were methylated divided by the number of clones sequenced (100 clones per dietary group; 25 clones per embryo pool from 1 mother, 4 mothers per group; n = 4 for statistical analyses). The data are presented in Fig. 3 and are plotted in two ways. The line graph maintains the distance relationships between the CpGs whose average methylation level is given on the y axis. The points are connected with a line not meant to imply continuity but to aid in readability. The inset shows the data presented as a “heat-map” of gray scale proportional to the methylation level of each CpG indicated. The plotting of the results as average methylation per dietary group at each CpG position showed a characteristic pattern (Fig. 3). For example, in the liver, CpGs at positions 8 and 15 were highly methylated (30-60%), whereas the 19th CpG was poorly methylated (4-8%). There was also a tendency for higher methylation in the core of DMR2 (positions 5-12). Surprisingly, but in agreement with the results on global DNA methylation reported above, choline deficiency was associated with the highest degree of methylation in the liver at E17 (p < 0.00001, ANOVA). The whole experiment was duplicated, and similar results were obtained (Fig. 3). We then calculated the overall DNA methylation level by averaging the methylation values for each CpG using the combined data from both experiments. Methylation levels in choline-deficient animals (35%) were significantly higher than in controls (20%) (p < 0.01; n = 8) (Fig. 4). Moreover, average methylation levels for the DMR2 core (mean methylation values of each dietary group for the positions 5-12) was lower in control animals (26%) than in choline-deficient (46%) and choline-supplemented subjects (34%) (p < 0.001; n = 8) (Fig. 4). With the exception of position 11, all individual positions in the core of DMR2 showed significant hypermethylation in the choline-deficient embryos as compared with controls. In addition, position 13, the CpG adjacent to the core of DMR2, and positions 1 and 15 outside of the core of DMR2 displayed DNA hypermethylation in choline-deficient animals (Fig. 3). Position 15 showed the highest level of methylation in all groups. In this position choline-deficient embryos had 66% methylation, significantly higher (p < 0.0005) than controls, which had 38% methylation (Fig. 3).
      Figure thumbnail gr3
      FIGURE 3Bisulfite sequencing analysis of Igf2 DMR2 in the liver of E17 rat embryos. The experimental samples were prepared as described under “Experimental Procedures.” Bisulfite-modified DNA was amplified using primers that generate a product spanning a 20 CpG-containing region of DMR2 of Igf2 and 25 clones per embryonic pool from one mother; a total 100 clones per group were sequenced as described under “Experimental Procedures.” The results are presented as average methylation in each CpG site per dietary group (n = 4 per experiment). The distances between the positions are proportionate to the actual distances on the chromosome. The lines connecting the points are drawn to improve the readability and are not meant to imply continuity. The inset shows the data as a heat map (a high level of methylation in black, no methylation in white). Choline deficiency in E17 liver is associated with hypermethylation of DMR2 as compared with the other groups (p < 0.00001; Tukey test). The data from two separate experiments are shown, and in both cases, the characteristic methylation pattern repeated with remarkable consistency.
      Figure thumbnail gr4
      FIGURE 4Bisulfite sequencing analysis of Igf2 DMR2 in the liver of E17 rat embryos; total methylation levels and positions sensitive to the effect of choline availability. The experimental samples were prepared and analyzed as described in . The methylation levels for the DMR2 and for the DMR2 core (positions 5-12, see ) were calculated by averaging the methylation levels of each dietary group from experiments 1 and 2 combined (left panel). The results were analyzed by ANOVA followed by a Tukey test. Choline-deficient animals had total and core methylation levels significantly higher than controls (p < 0.006 and p < 0.003, respectively). The average methylation of each embryo pool in each individual core CpG position was also analyzed by one-way ANOVA followed by a Tukey test (right panel), and seven of eight CpGs in the DMR2 core showed significant difference (p < 0.05); the choline-deficient animals had higher levels of methylation as compared with controls.
      Igf2 DMR2 and DMD Methylation Status in E17 Frontal Cortex—To analyze the DMR2 in frontal cortex, we used bisulfite PCR, cloning, and sequencing as described above. As in liver, most clones were partially methylated, and only a small fraction of the clones was fully unmethylated. Likewise, the total percentage of methylation per dietary group was calculated as above. The plotting of these results as the mean methylation per dietary group at each CpG position showed a characteristic pattern in the frontal cortex, which was similar to that observed in the liver (Fig. 5). The highest methylation was again observed in the core of DMR2 (positions 5-12); however, in contrast to liver, position 15 was not highly methylated. Moreover, the effect of the diet was also observed in the frontal cortex at E17 such that choline-supplemented embryos had lower degree of methylation as compared with the controls (p < 0.005). Individual CpG analysis of the average methylation of each embryo pool in each position by ANOVA revealed an effect of diet on positions 7 and 19. At position 7, choline-supplemented embryos had average methylation level of 10%, which was significantly lower than choline-deficient embryos, whose methylation level was 36% (Fig. 5). Interestingly, choline-deficient animals had 6 times higher level of methylation as compared with the controls in position 19 (p < 0.01) (Fig. 5). In addition, we investigated the methylation status of another Igf2 regulatory region called DMD. This region contains four subregions that, when unmethylated, bind the CTCF protein that represses Igf2 transcription. We used bisulfite-treated DNA from E17 frontal cortex and a primer pair designed to analyze the 334-bp product of DMD subregion r3 that contained the CTCF binding site and sequenced 10-15 clones per embryo pool from 1 mother, 4 mothers in each dietary group. The overall methylation level was ∼40%; most of the clones were fully methylated or fully unmethylated, and only a few clones showed partial methylation. The results demonstrate that in E17 DMD subregion r3 of the frontal cortex, there was no significant difference in DNA methylation between the groups (data not shown).
      Figure thumbnail gr5
      FIGURE 5Bisulfite sequencing analysis of Igf2 DMR2 in the frontal cortex of E17 rat embryos. The experimental samples were prepared as described in . Bisulfite-modified DNA was amplified using primers that generate a product spanning a 20 CpG-containing region of DMR2 of Igf2 and 25 clones per embryonic pool from 1 mother; a total of 100 clones per group were sequenced as described under “Experimental Procedures.” The results are presented as average methylation in each CpG site per dietary group (n = 4) as described in . E17 choline-supplemented embryos had the lowest degree of methylation in frontal cortex as compared with the control and choline-deficient animals (p < 0.005; two-way ANOVA followed by a Tukey test).
      Igf2 ExpressionIgf2 expression is regulated by methylation in liver in such a way that hypermethylation of either DMR2 or DMD leads to increased expression. Thus, we determined the levels of Igf2 transcript using RT-PCR. In E17 liver (Fig. 6), the level of Igf2 mRNA was lower by 50% in the choline-supplemented embryos as compared with control and choline-deficient embryos (p < 0.05 for each). These results are consistent with the current notion of the regulation of Igf2, according to which increased methylation in DMR2 results in increased Igf2 expression. Consistent with previous studies, the levels of Igf2 transcript in frontal cortex of E17 embryos were too low to be assayed reliably (
      • Stylianopoulou F.
      • Kim A.
      • Herbert J.
      • Pintar J.
      ,
      • Beck F.
      • Kim N.J.
      • Penschow J.D.
      • Thorley B.
      • Tregear G.W.
      • Coghlan J.P.
      ).
      Figure thumbnail gr6
      FIGURE 6Expression of Igf2 mRNA in E17 rat liver as determined by RT-PCR. Total RNA was purified and used for RT-PCR analysis as described under “Experimental Procedures.” The mRNA level for Igf2 was normalized to that of β-actin and expressed as percent of the control group. Data are presented as means ± S.E. Choline-supplemented embryos had lower Igf2 mRNA levels as compared with control and choline deficient animals (p < 0.05; Tukey test).
      Dnmt and Mbd2 Expression in E17 Liver and Frontal Cortex—One possible mechanism that leads to changes in the global as well as gene-specific DNA methylation is via alteration in the activity of DNMTs. It has been shown that DNMT1 is important for maintaining the methylation pattern of the Igf2 gene, and Dnmt1 knock-out mice had abnormal expression of Igf2 (
      • Li E.
      • Kim C.
      • Jaenisch R.
      ). Moreover, DMR2 methylation pattern is dependent on DNMT3A and DNMT3B activity in a process that is important for the establishment of the de novo methylation patterns in embryonic stem cells (
      • Okano M.
      • Kim D.W.
      • Haber D.A.
      • Li E.
      ). Thus, methylation of Igf2 can be influenced by all three DNMTs. We assayed DNMT expression using RT-PCR. In the liver of choline-deficient embryos, Dnmt1 (Fig. 7), the “maintenance” enzyme, was overexpressed by more than 50% as compared with control and choline-supplemented embryos (p < 0.05). Because we performed several assays on tissue extracts derived from the same animals, it was possible to perform a correlation analyses on our data. Interestingly, there was a significant correlation in E17 liver between the mRNA levels of Dnmt1 and DNA 5-methylcytosine content (r = 0.66, p < 0.02) (Fig. 8A). Moreover, there was a significant correlation in E17 liver between the mRNA levels of Dnmt1 and total methylation of all CpGs within the sequenced DMR2 region of Igf2 (r = 0.64, p < 0.03) (Fig. 8B), and there was a linear correlation between mRNA levels of Igf2 and Dnmt1 (r = 0.60, p < 0.03) (Fig. 8C). In addition, Dnmt3a (Fig. 9A), a de novo methyltransferase, mRNA levels were lower by ∼45% in the choline-supplemented embryos as compared with control and choline-deficient embryos (p < 0.05), whereas the levels of mRNA encoding MBD2 (Fig. 9B), a protein that binds to methylated CpGs and acts as a methylation-dependent transcriptional repressor with possible demethylase activity (
      • Ballestar E.
      • Kim A.P.
      ), were higher by 40% in choline-deficient as compared with control and choline-supplemented embryos (p < 0.05). The diet had no significant effect on the expression of the de novo methyltransferase Dnmt3b and the regulator of the de novo methyltransferases, Dnmt3l (data not shown). In the frontal cortex, Dnmt1 (Fig. 10A), Dnmt3a (Fig. 10B), and Mbd2 (Fig. 10D) mRNA levels were lower by 15, 30, and 30% respectively, in choline-supplemented embryos as compared with controls (for all, p < 0.05). Interestingly, in the frontal cortex of the choline-deficient embryos, Dnmt3l levels (Fig. 10C) were almost 2-fold higher than controls (p < 0.05). Taken together, the data point to the possibility that the observed changes in global and gene specific methylation (Figs. 2, 3, 4 and 5) result from the changes in the methyltransferase machinery (Figs. 7, 9, and 10).
      Figure thumbnail gr7
      FIGURE 7Expression of Dnmt1 mRNA in E17 rat liver as determined by RT-PCR. The experimental samples were prepared, analyzed, and presented as described in . Dnmt1 mRNA was overexpressed in liver of choline-deficient embryos, as compared with control and choline-supplemented embryos (p < 0.05; Tukey test).
      Figure thumbnail gr8
      FIGURE 8Correlation analysis of the relationship between Dnmt1 mRNA levels and DNA global methylation, Igf2 DMR2 total methylation levels, and Igf2 mRNA levels. The data for the correlation analyses are shown in Figs. , , and . There was a significant correlation in E17 liver between the mRNA levels of Dnmt1 and DNA 5-methylcytosine content (r = 0.66, p < 0.02) (A) and average methylation of all CpGs within the sequenced region of Igf2 DMR2 (r = 0.64, p < 0.03) (B) and mRNA levels of Igf2 (r = 0.60, p < 0.03) (C).
      Figure thumbnail gr9
      FIGURE 9Expression of Dnmt3a and Mbd2 in the liver of E17 embryos. The samples were prepared and analyzed as described under “Experimental Procedures.” The mRNA levels for each gene were normalized to that of β-actin and expressed as percentage of the control group. Data are presented as means ± S.E. A, choline supplementation was associated with lower levels of Dnmt3a mRNA as compared with controls (p < 0.05; Tukey test). B, in the liver of E17 embryos, Mbd2 was overexpressed in the choline-deficient group (p < 0.05; Tukey test) as compared with control and choline-supplemented embryos.
      Figure thumbnail gr10
      FIGURE 10Expression of DNA methylating and de-methylating enzymes in the frontal cortex of E17 embryos. The samples were prepared and analyzed as described under “Experimental Procedures.” The mRNA level for each gene was normalized to that of β-actin and expressed as a percentage of the control group. Data are presented as means ± S.E. A, choline supplementation was associated with lower levels of Dnmt1 (A), Dnmt3a (B), and Mbd2 (D) mRNA as compared with controls (p < 0.05; Tukey test). Dnmt3l was overexpressed in the choline-deficient group (p < 0.05; Tukey test) as compared with control and choline-supplemented embryos (C).
      Methylation of the Dnmt1 Gene in E17 Liver and Cortex—DNA methylation of a regulatory region of Dnmt1 modulates its expression in cis. Slack et al. (
      • Slack A.
      • Kim N.
      • Pinard M.
      • Szyf M.
      ) showed that hypomethylation of a single CpG of mouse Dnmt1 gene (designated as position 101 by these authors) up-regulates its expression. Comparison of the mouse and rat sequence showed that this CpG was conserved, and we used methylation-specific primer analysis to estimate the degree of its methylation in E17 rat tissues. We found that this CpG was significantly hypomethylated in the liver of choline-deficient embryos as compared with controls, whereas in the cerebral cortex this CpG was hypermethylated in choline-supplemented animals as compared with controls and choline-deficient fetuses (Fig. 11).
      Figure thumbnail gr11
      FIGURE 11Methylation-specific PCR of the regulatory CpG of the Dnmt1 gene in E17 liver and cerebral cortex. Bisulfite DNA was prepared from livers and cerebral cortex and methylation-specific PCR was performed as described under “Experimental Procedures.” The results shown are ratios of the amount of PCR product obtained using a primer specific for the methylated template relative to that obtained with a primer specific for the unmethylated template normalized to the control group (M:U). These values were statistically significantly lower in liver of choline-deficient fetuses relative to controls (p < 0.05; Tukey test) and statistically significantly higher in cortex of choline-supplemented fetuses relative to controls and choline-deficient animals (p < 0.005 and p < 0.006, respectively; Tukey test).

      DISCUSSION

      Taken together, our data show a remarkable adaptive response of the fetus to a relatively short period (E11-E17) of altered dietary choline intake by the pregnant mother, comprising changes in hepatic and cerebral DNA methylation and modified expression of multiple genes, including those encoding proteins that participate in the processes of DNA methylation. Surprisingly, we found that in the liver of E17 animals, choline deficiency leads to global and Igf2 gene hypermethylation, likely caused by an apparently compensatory induction of Dnmt1. Moreover, the latter may be due to the hypomethylation of a CpG in its regulatory region (
      • Slack A.
      • Kim N.
      • Pinard M.
      • Szyf M.
      ), observed in the choline-deficient fetuses. Thus, we posit that choline deficiency causes an initial hypomethylation of regulatory DNA sequences in Dnmt1 followed by induction of its expression leading to an overall acceleration of DNA methylation that manifests as increased levels of DNA 5-methylcytosine content. One would expect that this feedback mechanism may counteract the shortage of choline-derived methyl groups for some period of time; however, full compensation for a prolonged period of choline and/or methyl group deficiency is unlikely.
      We elected to study the effects of altered choline availability on liver and frontal cortex. The former was chosen because of its relative cellular uniformity, its key role in the metabolism of choline, and because previous literature showed that diet affects methylation in that organ in the adult (
      • Bhave M.R.
      • Kim M.J.
      • Poirier L.A.
      ,
      • James S.J.
      • Kim I.P.
      • Pogribna M.
      • Miller B.J.
      • Jernigan S.
      • Melnyk S.
      ,
      • Kim Y.I.
      • Kim I.P.
      • Basnakian A.G.
      • Miller J.W.
      • Selhub J.
      • James S.J.
      • Mason J.B.
      ,
      • Esfandiari F.
      • Kim R.
      • Cotterman R.F.
      • Pogribny I.P.
      • James S.J.
      • Miller J.W.
      ). The choice of frontal cortex is based on our previous investigations showing changes in the expression of multiple genes, including Igf2, in postnatal rats subjected to altered choline availability in utero (
      • Mellott T.J.
      • Kim M.T.
      • Diesl V.
      • Hill A.A.
      • Lopez-Coviella I.
      • Blusztajn J.K.
      ) and based on the report by Niculescu et al. (
      • Niculescu M.D.
      • Kim C.N.
      • Zeisel S.H.
      ) showing that a subpopulation of brain cells responds to choline deficiency by DNA hypomethylation in fetal mice. It is worth noting that there is a paucity of data on the effects of choline nutrition during pregnancy on fetal DNA methylation, and thus, by necessity, we must discuss our results in relation to previous work performed in postnatal animals.
      The levels of SAM and SAH measured in our experiments were consistent with those reported by others in postnatal rats (
      • Hoffman D.R.
      • Kim W.E.
      • Duerre J.A.
      ,
      • Gharib A.
      • Kim N.
      • Chabannes B.
      • Cronenberger L.
      • Pacheco H.
      ). In adult animals, choline- (
      • Zeisel S.H.
      • Kim T.
      • DaCosta K.-A.
      • Pomfret E.A.
      ) or methyl-deficient diets (
      • Hoffman D.R.
      • Kim J.A.
      • Cornatzer W.E.
      ) lowered hepatic SAM levels, and rats that were fed a diet supplemented with methionine, choline, folic acid, and vitamin B12 had increased hepatic levels of SAM (
      • Shivapurkar N.
      • Kim L.A.
      ). Similarly, the significantly higher levels of SAM in the liver and brain of choline-supplemented embryos shown here indicate that the dietary treatment was sufficient to influence methyl group metabolism. Previous investigations showed that choline consumed by pregnant rats is efficiently transported to the fetus (
      • Garner S.C.
      • Kim M.H.
      • Zeisel S.H.
      ), and thus, it is likely that the alterations in methyl group metabolism observed in this study were mediated by changes in the concentrations of fetal choline pools. However, it is also possible that the actions of choline are indirect and mediated, perhaps by changes in maternal methyl group metabolism that may include altered transport of methionine and/or SAM across the placenta. The fact that choline deficiency did not result in changes in SAM and SAH levels may be due to the short duration of our dietary intervention. Alternatively, the abundance of methyl group donors other than choline (e.g. methionine) might be sufficient to compensate, thereby preserving total tissue SAM and SAH levels in the fetal liver and brain. Because we assayed SAM and SAH in whole cell extracts, we cannot address the possibility that choline intake might have affected the levels of these compounds differentially in subcellular compartments. SAM is present in the cytoplasm, mitochondria, and the nucleus (
      • Farooqui J.Z.
      • Kim H.W.
      • Kim S.
      • Paik W.K.
      ), and the amount of nuclear SAM constitutes less than half of its total cellular pool (
      • Farooqui J.Z.
      • Kim H.W.
      • Kim S.
      • Paik W.K.
      ). We do not know if this nuclear SAM pool, which provides the substrate for DNMTs, was affected by choline intake.
      To determine whether choline availability in utero alters global DNA methylation, we measured the 5-methylcytosine content in liver and brain of embryos and found values that were similar to previous estimates of global DNA 5-methylcytosine content of ∼4 mol % (
      • Gehrke C.W.
      • Kim R.A.
      • Gama-Sosa M.A.
      • Ehrlich M.
      • Kuo K.C.
      ), and consistent with previous studies we observed higher levels of 5-methylcytosine in brain as compared with liver (
      • Tawa R.
      • Kim T.
      • Kurishita A.
      • Okada S.
      • Hirose S.
      ,
      • Ehrlich M.
      • Kim M.A.
      • Huang L.H.
      • Midgett R.M.
      • Kuo K.C.
      • McCune R.A.
      • Gehrke C.
      ). The effect of methyl group deficiency on global DNA methylation is complex. For example Niculescu et al. (
      • Niculescu M.D.
      • Kim C.N.
      • Zeisel S.H.
      ) found that global DNA methylation as well as methylation of the Cdkn3 gene, was reduced in microdissected ventricular and subventricular cells of the hippocampal Ammon's horn at E17 in mice whose mothers received a choline-deficient diet between E12 and E17 as compared with fetuses of control dams. Interestingly other brain areas were not affected. In our study, despite unaffected hepatic SAM and SAH levels, choline deficiency was associated with DNA hypermethylation in embryonic liver. Similarly, rats fed a moderately folate-deficient diet for 3 weeks had global DNA hypermethylation that did not correlate with SAM or SAH concentrations and was not observed beyond that time point (
      • Sohn K.J.
      • Kim J.M.
      • Reid S.
      • Shirwadkar S.
      • Mason J.B.
      • Kim Y.I.
      ). In addition, transient DNA hypermethylation was observed in the liver of mice fed a moderately folate-deficient diet for 5 weeks; however, DNA methylation returned to normal by the eighth week on the diet (
      • Song J.
      • Kim K.J.
      • Medline A.
      • Ash C.
      • Gallinger S.
      • Kim Y.I.
      ). Our data are also consistent with a recent report showing that in livers of E21 fetuses obtained from pregnant rats that were given a folate-deficient diet containing reduced amounts of choline and methionine periconceptually and throughout gestation, there was a significant DNA hypermethylation as compared with that in animals receiving a folate-deficient diet alone (
      • Maloney C.A.
      • Kim S.M.
      • Rees W.D.
      ).
      Most of the methylated DNA is located in parts of the genome not associated with genes. Therefore, we investigated the methylation status of two genomic regions known to regulate gene expression of Igf2 in a methylation-dependent manner, namely DMR2 and DMD. Igf2 is an imprinted gene, expressed from the paternal allele in most tissues, with the notable exception of the brain where it is expressed biallelically (
      • DeChiara T.M.
      • Kim E.J.
      • Efstratiadis A.
      ,
      • Hu J.F.
      • Kim T.H.
      • Hoffman A.R.
      ,
      • Pham N.V.
      • Kim M.T.
      • Hu J.F.
      • Vu T.H.
      • Hoffman A.R.
      ). Methylation of DMD and DMR2 of the expressed paternal allele tends to be high, whereas the silenced maternal allele is hypomethylated. Using bisulfite PCR, sequencing, and cloning, we observed a characteristic pattern in DMR2 methylation of Igf2 in E17 liver and cortex. Interestingly, a similar pattern of peaks and valleys of methylation levels of the equivalent murine CpGs was previously observed in mouse kidney (
      • Waterland R.A.
      • Kim J.R.
      • Smith C.A.
      • Jirtle R.L.
      ). Previous studies showed that high methylation of DMR2 correlates with high Igf2 expression in the liver (
      • Murrell A.
      • Kim S.
      • Bowden L.
      • Constancia M.
      • Dean W.
      • Kelsey G.
      • Reik W.
      ,
      • Feil R.
      • Kim J.
      • Allen N.D.
      • Reik W.
      ) and methylated DMR2 increased the expression of a reporter gene in a cell line (
      • Murrell A.
      • Kim S.
      • Bowden L.
      • Constancia M.
      • Dean W.
      • Kelsey G.
      • Reik W.
      ). Similarly, we find that hypermethylation of DMR2 in choline-deficient animals was associated with overexpression of Igf2 in embryonic liver. Surprisingly, almost all CpG positions in the core of DMR2 were hypermethylated in the choline-deficient animals as compared with controls. In addition, two positions outside of the core were similarly affected by choline availability, although the significance of the changes in those two positions is unknown. In our model we cannot address the parental origin of the sequenced clones, but it is possible that either the maternal, paternal, or both alleles become hypermethylated in choline deficiency. In studies by Waterland et al. (
      • Waterland R.A.
      • Kim J.R.
      • Smith C.A.
      • Jirtle R.L.
      ), mice fed an amino acid-defined synthetic diet for up to 60 days post-weaning showed allele-specific hypermethylation of paternal but not maternal DMR2 in the kidney.
      Analysis of the DMR2 of Igf2 in the frontal cortex revealed a methylation pattern fairly similar to the one seen in E17 liver. Note, however, that Igf2 mRNA is not expressed in fetal cortex. Overall, cortical DMR2 in E17 embryos was significantly hypomethylated in the choline-supplemented animals as compared with controls. When the methylation status at each of the CpG positions was analyzed individually, only two of those were significantly affected by the diet in the frontal cortex as compared with liver in which CpG methylation levels of 10 positions differed among the dietary groups. These data and the observations that DMR2, but not DMD, showed differences between the dietary cohorts indicate that specific regions of the genome, even within the same gene, are differentially susceptible to choline availability. Remarkably, however, our data obtained with assays of 5-methylcytosine content of the DNA as well as the degree of methylation of individual CpGs within DMR2 are internally consistent, showing an inverse relationship between maternal choline intake and DNA methylation status in E17 fetal liver and brain. This is particularly evident if one considers DMR2 methylation patterns in liver and cortex focusing only on the data from the choline-deficient and choline-supplemented groups (i.e. not including the control animals) (Figs. 3 and 5). Thus, it is the control group that resembles more the choline-supplemented one in the liver and the choline-deficient group in brain. One way to reconcile these data might be to propose that both liver and brain respond to choline deficiency by induction of Dnmt1 and subsequent DNA hypermethylation but that the liver turns on this mechanism when pregnant dams consume a choline-deficient diet, whereas the brain switches this system on in both choline-deficient rats and in animals fed the control diet. If this formulation is correct, then our control diet may be considered as providing adequate amounts of choline for hepatic, but not for cerebral, DNA methylation.
      DNA methylation is a dynamic process that depends on the supply of methyl groups, the activity of DNA methyltransferase enzymes, and DNA demethylase activity (
      • Ross S.A.
      ). Most of the studies of the effect of diet on DNMT expression were performed in a rat model of hepatocarcinogenesis using male Fischer rats. Thus, these rats weaned onto a methyl group-deficient diet for as little as 3 weeks displayed an increase in the expression and activity of cellular DNMT1, DNMT3A, and DNMT3B and a global DNA hypomethylation (
      • James S.J.
      • Kim I.P.
      • Pogribna M.
      • Miller B.J.
      • Jernigan S.
      • Melnyk S.
      ). In addition, in the same animal model, Dnmt1 and Dnmt3a mRNA levels were increased after 9 weeks on the methyl group donor-deficient diet (
      • Ghoshal K.
      • Kim X.
      • Datta J.
      • Bai S.
      • Pogribny I.
      • Pogribny M.
      • Huang Y.
      • Young D.
      • Jacob S.T.
      ,
      • Pogribny I.P.
      • Kim S.J.
      • Jernigan S.
      • Pogribna M.
      ). Similarly, we found overexpression of Dnmt1 in the liver of choline-deficient, as compared with choline-supplemented, and control embryos. Moreover, the levels of Dnmt1 mRNA correlated with global and Igf2 DMR2 DNA methylation. Because DNMT1 is known to catalyze not only methylation of hemimethylated templates but also to have some de novo methylation capacity (
      • Hermann A.
      • Kim H.
      • Jeltsch A.
      ), and because it is the most abundant cellular DNMT (
      • Hermann A.
      • Kim H.
      • Jeltsch A.
      ), it is possible that the observed changes in DNA methylation resulted from the increased expression of Dnmt1. A complementary trend was observed in the brain, where we noted a decreased expression of Dnmt1 in the frontal cortex of choline-supplemented embryos as compared with the controls and concomitant hypomethylation in the Igf2 DMR2. Thus, the comparison of Dnmt1 mRNA levels in liver and cortex provides evidence that Dnmt1 expression is inversely related to the intake of choline and suggests that it may be regulated in a similar fashion in these organs. Slack et al. (
      • Slack A.
      • Kim N.
      • Pinard M.
      • Szyf M.
      ) found a regulatory region in the Dnmt1 gene in which methylation of a single CpG (position 101) in close proximity to an Sp1 binding site results in decreased Dnmt1 expression. We, thus, determined the relative degree of methylation of this CpG and found hypomethylation at this site in the liver of choline-deficient rats and hypermethylation in cerebral cortex of choline-supplemented animals. Thus, we postulate that the methylation level of Dnmt1 regulatory site is proportional to the availability of choline, and in this fashion it may function as a choline (or methyl group) sensor. When choline is in short supply, the regulatory CpGs in Dnmt1 become hypomethylated, leading to the induction of Dnmt1 expression resulting in the increased rate of DNA methylation catalyzed by the enzyme. In contrast, when the supply of choline is high, these CpGs are hypermethylated, leading to the attenuation of Dnmt1 expression. For this formulation to be true, the regulatory CpGs of Dnmt1 must not serve as good substrates for DNMT1 to prevent feedback silencing of the gene by the induced DNMT1 enzyme.
      In addition to altering Dnmt1 expression, choline had similar effects in the liver and frontal cortex on the expression of Dnmt3a; that is, decreased mRNA levels in the choline-supplemented embryos as compared with the controls. In this case choline deficiency did not affect the expression of Dnmt3a, a result that contrasts with the data of others (
      • James S.J.
      • Kim I.P.
      • Pogribna M.
      • Miller B.J.
      • Jernigan S.
      • Melnyk S.
      ,
      • Ghoshal K.
      • Kim X.
      • Datta J.
      • Bai S.
      • Pogribny I.
      • Pogribny M.
      • Huang Y.
      • Young D.
      • Jacob S.T.
      ). It is interesting that choline availability affected the expression of Dnmt3l in the frontal cortex (Fig. 10C). The DNMT3L protein reportedly binds to DNMT3A2 and to histone H3 that is not methylated on lysine 4, thereby recruiting DNMT3A2 to DNA to permit DNA methylation by the latter (
      • Ooi S.K.
      • Kim C.
      • Bernstein E.
      • Li K.
      • Jia D.
      • Yang Z.
      • Erdjument-Bromage H.
      • Tempst P.
      • Lin S.P.
      • Allis C.D.
      • Cheng X.
      • Bestor T.H.
      ). To our knowledge this is the first report demonstrating its transcript in the brain as well as regulation of its expression by diet.
      Dietary factors can also influence DNA demethylation. To date the only known protein with a possible methylated-DNA demethylase activity is MBD2 (
      • Detich N.
      • Kim J.
      • Szyf M.
      ). MBD2 binds to methylated CpGs and acts as a transcriptional repressor (
      • Ballestar E.
      • Kim A.P.
      ). Rats fed a methyl group-deficient diet had increased hepatic expression of MBD2 (
      • Ghoshal K.
      • Kim X.
      • Datta J.
      • Bai S.
      • Pogribny I.
      • Pogribny M.
      • Huang Y.
      • Young D.
      • Jacob S.T.
      ,
      • Esfandiari F.
      • Kim R.
      • Cotterman R.F.
      • Pogribny I.P.
      • James S.J.
      • Miller J.W.
      ). Similarly, we observe an inverse relationship between the levels of Mbd2 mRNA expression and dietary choline intake in liver and brain. It has been proposed that MBD2 interacts with only certain promoters (
      • Detich N.
      • Kim J.
      • Szyf M.
      ), and thus, changes in its expression observed in our studies may result in regulation of the methylation status and/or expression of only limited number of genes.
      The epigenomic and apparently adaptive responses to gestational choline supply may initiate the long term developmental changes that characterize rats with varied choline intake in utero. Although it is possible that some of the changes in DNA methylation observed at E17 may be inherited through cell divisions occurring during subsequent developmental stages, it is also plausible that, as animals are returned to the control diet, the changes are erased. Alternatively, the altered pattern of gene expression during fetal life may modulate the trajectory of subsequent development without long-term preservation of the DNA methylation levels. As noted above, Igf2 is overexpressed in the liver of prenatally choline-deficient rats, and consistent with our previous studies (
      • Mellott T.J.
      • Kim M.T.
      • Diesl V.
      • Hill A.A.
      • Lopez-Coviella I.
      • Blusztajn J.K.
      ), Igf2 is not expressed in brain at E17, yet its expression in the cortex and hippocampus during the postnatal period is up-regulated in animals prenatally supplemented with choline (
      • Mellott T.J.
      • Kim M.T.
      • Diesl V.
      • Hill A.A.
      • Lopez-Coviella I.
      • Blusztajn J.K.
      ). Thus, we are unable at present to make predictions about the effects of prenatal choline availability on gene expression during postnatal life based on its epigenomic actions observed prenatally. Moreover, we have recently reported that prenatal choline deficiency up-regulates the expression of the choline transporter protein, CHT, in brain (
      • Mellott T.J.
      • Kim N.W.
      • Lopez-Coviella I.
      • Blusztajn J.K.
      ). CHT catalyzes the uptake of choline into cholinergic neurons, and thus, its long-term overexpression may be an adaptation to choline deficiency that is maintained even when the supply of the nutrient is returned to normal. Given that the rat CHT gene has a CpG island near its promoter (chr 9:133805-134277; UCSC Genome Browser), it is possible that prenatally acquired DNA methylation patterns may modulate its expression in adulthood.
      In the present study we concentrated on the effects of maternal choline deficiency and of moderate choline supplementation on fetal DNA methylation and gene expression with the overall aims of elucidating the molecular mechanisms of action of choline and of helping to define the adequate amount of dietary choline during pregnancy for normal fetal and postnatal development. Our data show that choline deficiency causes dramatic epigenetic changes and that the diet deemed as “control” may be marginal for choline content, whereas the supplemented diet contains enough choline not to evoke the epigenomic response and, thus, could be considered choline-sufficient. In contrast, in mouse models, maternal diets containing much higher amounts of methyl donors (at least twice that used in this study) promote DNA hypermethylation, presumably by increasing SAM levels, and lead to long term changes in gene expression and altered adult phenotypes in the offspring (
      • Wolff G.L.
      • Kim R.L.
      • Moore S.R.
      • Cooney C.A.
      ,
      • Waterland R.A.
      • Kim R.L.
      ,
      • Waterland R.A.
      • Kim M.
      • Tahiliani K.G.
      ,
      • Waterland R.A.
      • Kim D.C.
      • Lin J.R.
      • Smith C.A.
      • Shi X.
      • Tahiliani K.G.
      ,
      • Cropley J.E.
      • Kim C.M.
      • Beckman K.B.
      • Martin D.I.
      ). Recent studies in humans have brought about the realization that dietary requirements for choline vary and depend on the genotype due to polymorphisms in genes encoding enzymes involved in choline and folate metabolism (
      • Zeisel S.H.
      ). Converging data from studies of animal models and from epidemiological and clinical investigations indicate that individually adjusted appropriate choline intake during pregnancy is important for the normal development of the child and for adult health.

      References

        • Pitkin R.M.
        Am. J. Clin. Nutr. 2007; 85: 285S-288S
        • Blusztajn J.K.
        Science. 1998; 281: 794-795
        • Meck W.H.
        • Kim C.L.
        Neurosci. Biobehav. Rev. 2003; 27: 385-399
        • McCann J.C.
        • Kim M.
        • Ames B.N.
        Neurosci. Biobehav. Rev. 2006; 30: 696-712
        • Zeisel S.H.
        Annu. Rev. Nutr. 2006; 26: 229-250
        • Shaw G.M.
        • Kim S.L.
        • Yang W.
        • Selvin S.
        • Schaffer D.M.
        Am. J. Epidemiol. 2004; 160: 102-109
        • Meck W.H.
        • Kim C.L.
        Neuroreport. 1997; 8: 3045-3051
        • Mellott T.J.
        • Kim C.L.
        • Meck W.H.
        • Blusztajn J.K.
        FASEB J. 2004; 18: 412-427
        • Meck W.H.
        • Kim R.A.
        • Williams C.L.
        Dev. Psychobiol. 1988; 21: 339-353
        • Meck W.H.
        • Kim R.A.
        • Williams C.L.
        Behav. Neurosci. 1989; 103: 1234-1241
        • Meck W.H.
        • Kim C.L.
        Neuroreport. 1997; 8: 2831-2835
        • Meck W.H.
        • Kim C.L.
        Neuroreport. 1997; 8: 3053-3059
        • Meck W.H.
        • Kim C.L.
        Dev. Brain Res. 1999; 118: 51-59
        • Mellott T.J.
        • Kim M.T.
        • Diesl V.
        • Hill A.A.
        • Lopez-Coviella I.
        • Blusztajn J.K.
        FASEB J. 2007; 21: 1311-1323
        • Finkelstein J.D.
        • Kim J.J.
        • Harris B.J.
        • Kyle W.E.
        J. Nutr. 1983; 113: 519-521
        • Van den Veyver I.B.
        Annu. Rev. Nutr. 2002; 22: 255-282
        • Poirier L.A.
        • Kim P.H.
        • Rogers A.E.
        Cancer Res. 1977; 37: 744-748
        • Shivapurkar N.
        • Kim L.A.
        Carcinogenesis. 1983; 4: 1051-1057
        • Zeisel S.H.
        • Kim T.
        • DaCosta K.-A.
        • Pomfret E.A.
        Biochem. J. 1989; 259: 725-729
        • Bird A.
        Genes Dev. 2002; 16: 6-21
        • Bernstein B.E.
        • Kim A.
        • Lander E.S.
        Cell. 2007; 128: 669-681
        • Alonso-Aperte E.
        • Kim G.
        Int. J. Vitam. Nutr. Res. 1996; 66: 232-236
        • Wainfan E.
        • Kim M.
        • Stender M.
        • Christman J.K.
        Cancer Res. 1989; 49: 4094-4097
        • Wilson M.J.
        • Kim N.
        • Poirier L.A.
        Biochem. J. 1984; 218: 987-990
        • Christman J.K.
        • Kim G.
        • Dizik M.
        • Abileah S.
        • Wainfan E.
        Carcinogenesis. 1993; 14: 551-557
        • Niculescu M.D.
        • Kim C.N.
        • Zeisel S.H.
        FASEB J. 2006; 20: 43-49
        • Bhave M.R.
        • Kim M.J.
        • Poirier L.A.
        Carcinogenesis. 1988; 9: 343-348
        • James S.J.
        • Kim I.P.
        • Pogribna M.
        • Miller B.J.
        • Jernigan S.
        • Melnyk S.
        J. Nutr. 2003; 133: 3740S-3747S
        • Kim Y.I.
        • Kim I.P.
        • Basnakian A.G.
        • Miller J.W.
        • Selhub J.
        • James S.J.
        • Mason J.B.
        Am. J. Clin. Nutr. 1997; 65: 46-52
        • Kim Y.I.
        • Kim J.K.
        • Fleet J.C.
        • Cravo M.L.
        • Salomon R.N.
        • Smith D.
        • Ordovas J.
        • Selhub J.
        • Mason J.B.
        Am. J. Clin. Nutr. 1995; 61: 1083-1090
        • Song J.
        • Kim K.J.
        • Medline A.
        • Ash C.
        • Gallinger S.
        • Kim Y.I.
        Cancer Res. 2000; 60: 3191-3199
        • Sohn K.J.
        • Kim J.M.
        • Reid S.
        • Shirwadkar S.
        • Mason J.B.
        • Kim Y.I.
        Carcinogenesis. 2003; 24: 81-90
        • Hsu D.W.
        • Kim M.J.
        • Lee T.L.
        • Wen S.C.
        • Chen X.
        • Shen C.K.
        Proc. Natl. Acad. Sci. U. S. A. 1999; 96: 9751-9756
        • Okano M.
        • Kim E.
        J. Nutr. 2002; 132: 2462S-2465S
        • Singal R.
        • Kim J.M.
        Blood. 2001; 98: 3441-3446
        • Ghoshal K.
        • Kim X.
        • Datta J.
        • Bai S.
        • Pogribny I.
        • Pogribny M.
        • Huang Y.
        • Young D.
        • Jacob S.T.
        J. Nutr. 2006; 136: 1522-1527
        • Nan X.
        • Kim H.H.
        • Johnson C.A.
        • Laherty C.D.
        • Turner B.M.
        • Eisenman R.N.
        • Bird A.
        Nature. 1998; 393: 386-389
        • Jones P.L.
        • Kim G.J.
        • Wade P.A.
        • Vermaak D.
        • Kass S.U.
        • Landsberger N.
        • Strouboulis J.
        • Wolffe A.P.
        Nat. Genet. 1998; 19: 187-191
        • Eden S.
        • Kim T.
        • Keshet I.
        • Cedar H.
        • Thorne A.W.
        Nature. 1998; 394: 842
        • Takizawa T.
        • Kim K.
        • Namihira M.
        • Ochiai W.
        • Uemura A.
        • Yanagisawa M.
        • Fujita N.
        • Nakao M.
        • Taga T.
        Dev. Cell. 2001; 1: 749-758
        • Murrell A.
        • Kim S.
        • Bowden L.
        • Constancia M.
        • Dean W.
        • Kelsey G.
        • Reik W.
        EMBO Rep. 2001; 2: 1101-1106
        • Eden S.
        • Kim M.
        • Hashimshony T.
        • Dean W.
        • Goldstein B.
        • Johnson A.C.
        • Keshet I.
        • Reik W.
        • Cedar H.
        EMBO J. 2001; 20: 3518-3525
        • Bell A.C.
        • Kim G.
        Nature. 2000; 405: 482-485
        • Bondy C.A.
        • Kim H.
        • Roberts Jr., C.T.
        • LeRoith D.
        Mol. Endocrinol. 1990; 4: 1386-1398
        • Bieri J.G.
        • Kim G.S.
        • Briggs G.M.
        • Phillips R.W.
        • Woodard J.C.
        • Kanapka J.J.
        J. Nutr. 1977; 107: 1340-1348
        • Bieri J.G.
        J. Nutr. 1980; 110: 1726
        • Zeisel S.H.
        • Kim M.H.
        • Zhou Z.W.
        • Da Costa K.A.
        J. Nutr. 1995; 125: 3049-3054
        • Chomczynski P.
        • Kim N.
        Anal. Biochem. 1987; 162: 156-159
        • Molloy A.M.
        • Kim D.G.
        • Kennedy G.
        • Kennedy S.
        • Scott J.M.
        Biomed. Chromatogr. 1990; 4: 257-260
        • Gehrke C.W.
        • Kim R.A.
        • Gama-Sosa M.A.
        • Ehrlich M.
        • Kuo K.C.
        J. Chromatogr. 1984; 301: 199-219
        • Pruitt K.
        • Kim A.S.
        • Frantz K.
        • Rojas R.J.
        • Muniz-Medina V.M.
        • Rangnekar V.M.
        • Der C.J.
        • Shields J.M.
        J. Biol. Chem. 2005; 280: 23363-23370
        • Esfandiari F.
        • Kim R.
        • Cotterman R.F.
        • Pogribny I.P.
        • James S.J.
        • Miller J.W.
        Carcinogenesis. 2003; 24: 1935-1940
        • Grunau C.
        • Kim S.J.
        • Rosenthal A.
        Nucleic Acids Res. 2001; 29 (65): 65
        • Frommer M.
        • Kim L.E.
        • Millar D.S.
        • Collis C.M.
        • Watt F.
        • Grigg G.W.
        • Molloy P.L.
        • Paul C.L.
        Proc. Natl. Acad. Sci. U. S. A. 1992; 89: 1827-1831
        • Li L.C.
        • Kim R.
        Bioinformatics. 2002; 18: 1427-1431
        • Bock C.
        • Kim S.
        • Mikeska T.
        • Paulsen M.
        • Walter J.
        • Lengauer T.
        Bioinformatics. 2005; 21: 4067-4068
        • Shen L.
        • Kim Y.
        • Chen X.
        • Ahmed S.
        • Issa J.P.
        Biotechniques. 2007; 42 (50, and 52): 48
        • Lopes S.
        • Kim A.
        • Hajkova P.
        • Dean W.
        • Oswald J.
        • Forne T.
        • Murrell A.
        • Constancia M.
        • Bartolomei M.
        • Walter J.
        • Reik W.
        Hum. Mol. Genet. 2003; 12: 295-305
        • Stylianopoulou F.
        • Kim A.
        • Herbert J.
        • Pintar J.
        Development. 1988; 103: 497-506
        • Beck F.
        • Kim N.J.
        • Penschow J.D.
        • Thorley B.
        • Tregear G.W.
        • Coghlan J.P.
        Development. 1987; 101: 175-184
        • Li E.
        • Kim C.
        • Jaenisch R.
        Nature. 1993; 366: 362-365
        • Okano M.
        • Kim D.W.
        • Haber D.A.
        • Li E.
        Cell. 1999; 99: 247-257
        • Ballestar E.
        • Kim A.P.
        Eur. J. Biochem. 2001; 268: 1-6
        • Slack A.
        • Kim N.
        • Pinard M.
        • Szyf M.
        Eur. J. Biochem. 1999; 264: 191-199
        • Hoffman D.R.
        • Kim W.E.
        • Duerre J.A.
        Can J. Biochem. 1979; 57: 56-65
        • Gharib A.
        • Kim N.
        • Chabannes B.
        • Cronenberger L.
        • Pacheco H.
        J. Neurochem. 1982; 38: 810-815
        • Hoffman D.R.
        • Kim J.A.
        • Cornatzer W.E.
        Can. J. Biochem. 1981; 59: 543-550
        • Garner S.C.
        • Kim M.H.
        • Zeisel S.H.
        J. Nutr. 1995; 125: 2851-2858
        • Farooqui J.Z.
        • Kim H.W.
        • Kim S.
        • Paik W.K.
        Biochim. Biophys. Acta. 1983; 757: 342-351
        • Tawa R.
        • Kim T.
        • Kurishita A.
        • Okada S.
        • Hirose S.
        Differentiation. 1990; 45: 44-48
        • Ehrlich M.
        • Kim M.A.
        • Huang L.H.
        • Midgett R.M.
        • Kuo K.C.
        • McCune R.A.
        • Gehrke C.
        Nucleic Acids Res. 1982; 10: 2709-2721
        • Maloney C.A.
        • Kim S.M.
        • Rees W.D.
        Br. J. Nutr. 2007; 97: 1090-1098
        • DeChiara T.M.
        • Kim E.J.
        • Efstratiadis A.
        Cell. 1991; 64: 849-859
        • Hu J.F.
        • Kim T.H.
        • Hoffman A.R.
        Mol. Endocrinol. 1995; 9: 628-636
        • Pham N.V.
        • Kim M.T.
        • Hu J.F.
        • Vu T.H.
        • Hoffman A.R.
        Brain Res. 1998; 810: 1-8
        • Waterland R.A.
        • Kim J.R.
        • Smith C.A.
        • Jirtle R.L.
        Hum. Mol. Genet. 2006; 15: 705-716
        • Feil R.
        • Kim J.
        • Allen N.D.
        • Reik W.
        Development. 1994; 120: 2933-2943
        • Ross S.A.
        Ann. N. Y. Acad. Sci. 2003; 983: 197-207
        • Pogribny I.P.
        • Kim S.J.
        • Jernigan S.
        • Pogribna M.
        Mutat. Res. 2004; 548: 53-59
        • Hermann A.
        • Kim H.
        • Jeltsch A.
        Cell Mol. Life Sci. 2004; 61: 2571-2587
        • Ooi S.K.
        • Kim C.
        • Bernstein E.
        • Li K.
        • Jia D.
        • Yang Z.
        • Erdjument-Bromage H.
        • Tempst P.
        • Lin S.P.
        • Allis C.D.
        • Cheng X.
        • Bestor T.H.
        Nature. 2007; 448: 714-717
        • Detich N.
        • Kim J.
        • Szyf M.
        J. Biol. Chem. 2002; 277: 35791-35794
        • Mellott T.J.
        • Kim N.W.
        • Lopez-Coviella I.
        • Blusztajn J.K.
        Brain Res. 2007; 1151: 1-11
        • Wolff G.L.
        • Kim R.L.
        • Moore S.R.
        • Cooney C.A.
        FASEB J. 1998; 12: 949-957
        • Waterland R.A.
        • Kim R.L.
        Mol. Cell. Biol. 2003; 23: 5293-5300
        • Waterland R.A.
        • Kim M.
        • Tahiliani K.G.
        FASEB J. 2007; (in press)
        • Waterland R.A.
        • Kim D.C.
        • Lin J.R.
        • Smith C.A.
        • Shi X.
        • Tahiliani K.G.
        Genesis. 2006; 44: 401-406
        • Cropley J.E.
        • Kim C.M.
        • Beckman K.B.
        • Martin D.I.
        Proc. Natl. Acad. Sci. U. S. A. 2006; 103: 17308-17312